JPWO2013042260A1 - Driving assistance device - Google Patents

Abstract

Based on the information on the detected moving body, including a plurality of model candidates that define the correspondence between the relative position between the moving body and the own vehicle detected around the own vehicle and the driving operation of the driver A model to be used is determined from a plurality of model candidates (S101), and driving assistance is executed (S113) based on the determined model and the driving operation (S104) of the driver after the moving body is detected. It is preferable that the model determined based on the determined model and the driving operation of the driver after the moving body is detected can be updated.

Description

  The present invention relates to a driving support device.

  Conventionally, a technique for recognizing a pedestrian is known. For example, in Patent Document 1, when a pedestrian is detected from an input image captured by an infrared camera, deceleration control for decelerating the vehicle speed to a predetermined speed by a brake operation or the like, or the presence of a pedestrian is indicated by a lamp, buzzer, or speaker. A technique for performing alarm control to be notified by voice from is disclosed.

JP 2005-196590 A

  Here, the reaction to the pedestrian is different for each driver. If support is made uniformly based on recognized pedestrian information, some drivers may feel uncomfortable. It is desirable to be able to provide driving assistance that matches the driver's feeling while suppressing the driver from feeling uncomfortable.

  An object of the present invention is to provide a driving assistance device that can perform driving assistance while suppressing the driver from feeling uncomfortable.

  The driving support apparatus of the present invention includes a plurality of model candidates that define information on the relative position between the moving body detected in the vicinity of the host vehicle and the host vehicle, and a correspondence relationship between the driving operation of the driver, A model to be used is determined from the plurality of model candidates based on information on the detected moving body, and driving is performed based on the determined model and the driving operation of the driver after the moving body is detected. It is characterized by performing support.

  In the driving support apparatus, it is preferable that the determined model can be updated based on the determined model and the driving operation of the driver after the moving body is detected.

  In the driving support device, the degree of compatibility between the determined model and the driving operation of the driver after the moving body is detected is calculated based on the correspondence between the model and the driving operation for a predetermined number of times. However, it is preferable that the update is executed when the fitness is less than a set reference value.

  In the driving support apparatus, it is preferable that the determined model is updated according to a short-term fitness and a long-term fitness with the driving operation of the driver after the moving body is detected.

  In the driving support apparatus, the driving support is preferably based on a degree of deviation between the determined model and the driving operation of the driver after the moving body is detected.

  A driving support apparatus according to the present invention includes a plurality of model candidates that define correspondence relationships between information about the relative positions of a moving body and a host vehicle detected around the host vehicle and a driver's driving operation. A model to be used is determined from a plurality of model candidates based on the information related to the moving body, and driving assistance is executed based on the determined model and the driving operation of the driver after the moving body is detected. Therefore, the driving assistance apparatus according to the present invention has an effect that driving assistance can be performed while suppressing the driver from feeling uncomfortable.

FIG. 1 is a flowchart illustrating the operation of the driving support apparatus according to the embodiment. FIG. 2 is a diagram illustrating functions of the driving support apparatus according to the embodiment. FIG. 3 is a block diagram of the driving support apparatus according to the embodiment. FIG. 4 is a diagram illustrating a tension driving model. FIG. 5 is a diagram illustrating a standard operation model. FIG. 6 is a diagram illustrating a margin operation model. FIG. 7 is an explanatory diagram of a predicted side passage distance. FIG. 8 is an explanatory diagram of the deceleration rate. FIG. 9 is a diagram illustrating a target vehicle speed range. FIG. 10 is a diagram illustrating an example of a decision tree for model selection. FIG. 11 is a flowchart showing the model update operation. FIG. 12 is a diagram illustrating an example of the fitness calculation. FIG. 13 is a diagram illustrating an example of a model shift by the model update determination unit. FIG. 14 is an explanatory diagram of the degree of departure and the degree of departure recognition. FIG. 15 is a diagram illustrating an example of the number of data necessary for model update. FIG. 16 is a diagram illustrating a front crossing operation model. FIG. 17 is a diagram illustrating an example of an operation model in which the vertical axis indicates the operation timing.

  Hereinafter, a driving support apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings. In addition, this invention is not limited by this embodiment. In addition, constituent elements in the following embodiments include those that can be easily assumed by those skilled in the art or those that are substantially the same.

[Embodiment]
The embodiment will be described with reference to FIGS. 1 to 16. The present embodiment relates to a driving support device. FIG. 1 is a flowchart showing the operation of the driving support apparatus according to the embodiment, FIG. 2 is a diagram showing functions of the driving support apparatus according to the embodiment, and FIG. 3 is a block diagram of the driving support apparatus according to the embodiment. .

  The driving support apparatus 1-1 of the present embodiment models the driver's response to the posture and movement of the pedestrian, and whether or not the driver's response deviates from the standard based on the modeled result. Determine. The driving assistance apparatus 1-1 performs driving assistance when the difference between the driver's reaction and the modeled reference reaction is large or when the difference is predicted to be large. Therefore, according to the driving assistance apparatus 1-1 of the present embodiment, driving assistance can be executed based on the driver's reaction to the pedestrian, and driving assistance is provided while suppressing the driver from feeling uncomfortable. be able to.

  As shown in FIG. 2, the driving assistance apparatus 1-1 of this embodiment has a driving characteristic estimation function and a driving assistance function. The driving characteristic estimation function estimates the driving characteristic of the driver with respect to the object. Here, the object is a moving body around the host vehicle, for example, a moving body in front of the host vehicle. The moving body includes a pedestrian, a light vehicle such as a two-wheeled vehicle, and other objects that move on a road. The driving support apparatus 1-1 has a default driving behavior standard for an object created in advance. Prior to sufficient sampling to estimate the driver's driving characteristics, driving assistance is provided based on default driving behavior criteria. The driving characteristic estimation function can estimate the driving characteristic based on the driver's actual driving operation and can update the driving action standard.

  The driving support function performs driving support based on the driving action standard. The driving support function predicts the difference between the driving behavior standard and the actual driving operation of the driver, and determines whether or not to perform driving support and the support level of driving support. The driving assistance apparatus 1-1 according to the present embodiment performs driving assistance based on not only information related to a moving body such as a pedestrian but also a driving operation of the driver. If driving assistance is made uniformly based on information about a moving object, driving assistance that does not match the driver's feeling may occur. For example, a driver with a high skill level may feel that the assistance is excessive or troublesome for the same driving assistance. On the other hand, a driver with low skill may desire a higher level of assistance.

  The driving support apparatus 1-1 of the present embodiment can perform driving support in consideration of the driver's reaction to the posture and movement of a pedestrian or the like by performing driving support based on an actual driving operation. By determining whether or not to provide support based on the reaction to the moving body and the support level, driving support that matches the driver's feeling is possible. In addition, the support level is determined based on the driving operation, so that the driver who performs the driving operation deviating from the normality is informed of the presence of the pedestrian, etc. The level can be determined.

  As shown in FIG. 3, the driving support device 1-1 includes an object information calculation unit 10, a model database 11, a host vehicle information collection unit 12, a model selection unit 13, a model update determination unit 14, a model determination unit 15, and a driving. A behavior prediction unit 16, a driving behavior prediction determination unit 17, a support determination unit 18, a notice support unit 19, a vehicle control support unit 20, and a notice device 30 are provided.

  The object information calculation unit 10 calculates information related to the moving object as the object. In the following description, a case where the moving body is a pedestrian will be described as an example. The object information calculation unit 10 acquires information about pedestrians based on the detection results of various outside-vehicle environment sensors. The vehicle exterior environment sensor is, for example, a millimeter wave radar or a camera. The object information calculation unit 10 calculates pedestrian position information, pedestrian posture information, pedestrian behavior information, pedestrian attribute information, and the like based on the detection result of the outside environment sensor. The position information of the pedestrian includes the relative position of the pedestrian with respect to the own vehicle and the relative position of the pedestrian with respect to the lane in which the own vehicle travels. The pedestrian's posture information includes the orientation of the pedestrian's upper body part, the orientation of the pedestrian's face, and the pedestrian's posture (upright, forward leaning posture, etc.). The pedestrian behavior information includes the traveling direction of the pedestrian and the moving speed of the pedestrian. The attribute information of the pedestrian includes the age, sex, clothes, and occupation of the pedestrian. The calculation result of the object information calculation unit 10 is sent to the model selection unit 13.

  The own vehicle information collecting unit 12 collects information related to the own vehicle. Specifically, the host vehicle information collection unit 12 acquires the position of the host vehicle, the speed of the host vehicle, the steering angle of the host vehicle, the accelerator operation amount, the brake operation amount, the handle operation amount, and the like. A signal indicating the information collected by the own vehicle information collecting unit 12 is sent to the model selecting unit 13.

  The model selection unit 13 selects an operation model based on the object information. The model database 11 stores a plurality of models. The model selection unit 13 determines a driving model to be used for control from among the models stored in the model database 11 based on pedestrian characteristics such as the pedestrian's posture and behavior.

  More specifically, the model selecting unit 13 is a pedestrian (see reference numeral 42 in FIG. 8) between the reference measurement trigger time (when passing P0 in FIG. 8) and the measurement trigger time (when passing P1 in FIG. 8). , And the pedestrian's (a) position (within / outside a certain distance from the vehicle lane), (b) speed (steady / unsteady), (c) direction of travel (Crossing / translation), (d) posture (upright / walking), (e) posture direction (road direction / others), (f) upper body part direction (with / without vehicle direction confirmation), etc. Is selected.

  The operation model shown in FIGS. 4 to 6 is an example of a model stored in the model database 11. 4 is a diagram showing a tension operation model, FIG. 5 is a diagram showing a standard operation model, and FIG. 6 is a diagram showing a margin operation model. The driving model shown in FIGS. 4 to 6 is an example of a plurality of model candidates that define the correspondence between the information about the relative position between the moving body and the own vehicle detected around the own vehicle and the driving operation of the driver. It is. 4 to 6, the horizontal axis represents the predicted lateral passage distance, and the vertical axis represents the deceleration rate.

  FIG. 7 is an explanatory diagram of a predicted side passage distance. The predicted lateral passage distance is a predicted value of the distance W between the host lane 40 and the pedestrian 42 when the host vehicle 100 passes the position Pw on the host lane 40 corresponding to the position of the target pedestrian 42. . In other words, the side passing predicted distance is a predicted value of the interval W between the pedestrian 42 and the own lane 40 when passing the target pedestrian 42 as seen from the side. The distance W between the pedestrian 42 and the own lane 40 can be, for example, the size of the gap between the white line 41 on the sidewalk side of the own lane 40 and the pedestrian 42. The interval W between the pedestrian 42 and the own lane 40 is not limited to this, and may be an interval between the curbstone and the pedestrian 42, for example. That is, the predicted side-passing distance is a predicted value of the distance between the reference line or reference point of the own lane 40 and the pedestrian 42 when the host vehicle 100 passes by the pedestrian 42. In addition, the side passage predicted distance may be the size of the interval between the host vehicle 100 and the pedestrian 42. The predicted side-passing distance corresponds to the relative position between the moving body detected in the vicinity of the host vehicle and the host vehicle. The relative position is not limited to the predicted lateral passage distance.

  The deceleration rate is the deceleration rate of the host vehicle 100 in a predetermined section before the pedestrian 42 in the host lane 40. FIG. 8 is an explanatory diagram of the deceleration rate, and FIG. 9 is a diagram showing the target vehicle speed range. As shown in FIG. 8, the first point P0 and the second point P1 on the own lane 40 are determined based on the relative distance from the pedestrian 42 that is the object. A deceleration rate of the host vehicle 100 in a section between the first point P0 and the second point P1 is calculated.

The vehicle speed V0 of the host vehicle 100 is measured with the host vehicle 100 reaching the first point P0 as a reference measurement trigger. This vehicle speed V0 is also referred to as “reference vehicle speed V0”. The driving support apparatus 1-1 monitors the speed of the host vehicle 100 while traveling from the first point P0 to the second point P1, and stores the minimum value of the vehicle speed during this period as the minimum host vehicle speed V1. The deceleration rate is calculated using the measurement trigger as the arrival of the host vehicle 100 at the second point P1. The deceleration rate is calculated by the following formula (1).
Deceleration rate = 100 × {1- (V1 / V0)} (1)

  When the reference host vehicle speed V0 is a vehicle speed outside the target vehicle speed range, the minimum host vehicle speed V1 is not measured and the deceleration rate is not calculated. As shown in FIG. 9, the target vehicle speed range is defined as a vehicle speed range from the minimum vehicle speed Vmin to the maximum vehicle speed Vmax. The minimum vehicle speed Vmin is determined as a vehicle speed at which it can be estimated that the vehicle is traveling at a sufficiently low speed, for example. The maximum vehicle speed Vmax is determined, for example, as a vehicle speed at which the collision time TTC at the first point P0 is less than or equal to a certain value.

  Thus, the predicted side passage distance is based on information about a moving body such as a pedestrian, and the deceleration rate indicates the driving operation of the driver. Therefore, the driving model shown in FIGS. 4 to 6 is a model that defines the correspondence between the information about the moving body and the driving operation.

  As shown in FIGS. 4 to 6, high risk areas R1, R2, and R3, reference areas S1, S2, and S3 and low risk areas T1, T2, and T3 are set in each model. The reference areas S1, S2, and S3 are areas indicating the width of the deceleration rate that serves as a reference for the predicted lateral passage distance. The reference regions S1, S2, and S3 are determined based on a probability distribution when the deceleration rate is a random variable, for example. The reference region S1, S2, S3 of the default operation model is determined based on, for example, deceleration rate data obtained from experimental results or the like. The reference areas S1, S2, and S3 are defined as areas including a certain percentage of data including the central value data among all acquired data. Further, the reference areas S1, S2, and S3 are updated based on the deceleration rate generated by the past driving operation by the driver, as will be described later.

  Areas with a reduction rate more than the reference areas S1, S2, S3 are high risk areas R1, R2, R3. The high-risk areas R1, R2, and R3 are areas that can be predicted to increase the risk in the relationship between the host vehicle 100 and the pedestrian 42. For example, the host vehicle 100 approaches the pedestrian 42, and the host vehicle 100 is a pedestrian. This is an area where it is predicted that there is a high possibility that a sufficient interval cannot be maintained with respect to 42. The high-risk areas R1, R2, and R3 include areas where the deceleration rate is negative, that is, a case where acceleration is performed without deceleration between the first point P0 and the second point P1. High risk side boundary lines H1, H2, and H3, which are boundaries between the high risk areas R1, R2, and R3 in the reference areas S1, S2, and S3, are deceleration lines when the reference vehicle speed V0 is the minimum vehicle speed Vmin. is there. The high risk side boundary lines H1, H2, and H3 may be curved lines.

  The regions with a higher deceleration rate than the reference regions S1, S2, S3 are the low risk regions T1, T2, T3. The low risk side boundary lines L1, L2, and L3, which are boundaries between the low risk areas T1, T2, and T3 in the reference areas S1, S2, and S3, are deceleration straight lines when the reference vehicle speed V0 is the maximum vehicle speed Vmax. The low risk side boundary lines L1, L2, and L3 may be curved lines.

  The tension driving model shown in FIG. 4 is a driving model under a situation where the driver feels a relatively high tension. The tension driving model is selected, for example, when the distance from the lane 40 where the host vehicle 100 travels to the pedestrian 42 is small.

  The margin driving model shown in FIG. 6 is a driving model under a situation where the driver can deal with a margin without feeling so nervous. The margin driving model is selected, for example, when the pedestrian 42 stands away from the own lane 40 and faces the opposite side to the own lane 40 side.

  The standard operation model shown in FIG. 5 is an intermediate operation model between a tension operation model and a margin operation model. In other words, the standard driving model is a driving model under a situation where a moderate tension is felt.

  FIG. 10 is a diagram illustrating an example of a decision tree for model selection. For example, the model selection unit 13 of the present embodiment selects a model according to the decision tree of FIG. The model selection is performed when a pedestrian 42 is detected in front of the host vehicle 100. For example, each time the pedestrian 42 is detected by the object information calculation unit 10, information on the pedestrian 42 is selected. A model is selected based on When a plurality of pedestrians 42 are detected, a model may be selected for each pedestrian 42, and the model with the highest tension among the selected models may be used for control.

  In the decision tree, first, cases are classified according to the position of the pedestrian 42. The model selection unit 13 determines whether the pedestrian 42 is outside the lane 40 and within a certain distance from the lane 40. When the pedestrian 42 is within a certain distance from the own lane 40, the tension driving model is selected.

  When the pedestrian 42 is not within a certain distance from the own lane 40, cases are classified according to the posture of the pedestrian 42. The model selection unit 13 determines whether the pedestrian 42 is upright or walking. When it is determined that the posture of the pedestrian 42 is upright, case classification is performed based on the orientation of the posture of the pedestrian 42. On the other hand, when it is determined that the pedestrian 42 is walking, the cases are classified according to the traveling direction of the pedestrian 42.

  In the case classification according to the orientation of the pedestrian 42, it is determined whether the pedestrian 42 is facing the own lane 40 side or the opposite side (outside) from the own lane 40. The model selection unit 13 selects the standard driving model when it is determined that the pedestrian 42 is facing the own lane 40 side, and selects the margin driving model when it is determined that the pedestrian 42 is facing the outside. .

  In the case classification according to the traveling direction of the pedestrian 42, it is determined whether the traveling direction of the pedestrian 42 is a direction crossing the own lane 40 or a direction parallel to the own lane 40. When the model selection unit 13 determines that the traveling direction of the pedestrian 42 is a direction crossing the own lane 40, the model selection unit 13 selects a standard driving model. On the other hand, if the model selection unit 13 determines that the traveling direction is a direction parallel to the own lane 40, the model selecting unit 13 classifies the cases according to the speed of the pedestrian 42.

  In the case classification based on the speed, it is determined whether the moving speed of the pedestrian 42 is a steady speed or an unsteady speed. The model selection unit 13 selects a surplus driving model when the moving speed of the pedestrian 42 is a steady speed, and selects a standard driving model when the moving speed is an unsteady speed. Note that a corresponding model can be selected from a plurality of models in the same manner for moving objects other than pedestrians.

  The case elements for selecting a model are not limited to those shown in the figure. For example, cases may be classified according to the orientation of the upper body part of the pedestrian 42. When the direction of the upper body part is a direction for confirming the direction of the host vehicle 100, a model with a relatively low degree of tension is selected, and when not, a model with a relatively high degree of tension is selected. Good.

  The model update determination unit 14 performs update processing of the model selected by the model selection unit 13. The model update determination unit 14 can update the determined model based on the model determined to be used and the driving operation of the driver after the moving body is detected. FIG. 11 is a flowchart showing the model update operation. The model update determination unit 14 updates the model according to the flowchart shown in FIG. 11, for example. The flowchart shown in FIG. 11 is executed when a model is selected by the model selection unit 13.

  In step S201, the model update determination unit 14 calculates the fitness. The goodness of fit indicates the degree of goodness of the selected model with respect to the driving characteristics of the driver. The degree of adaptation indicates the degree of adaptation between the model determined to be used and the driving operation of the driver after the moving body is detected. The model update determination unit 14 includes a short-term update determination unit 14a that performs a short-term update based on a short-term fitness and a long-term update determination unit 14b that performs a long-term update based on a long-term fitness.

Short-term updates are based on the most recently specified number of samples. For example, when the currently selected model is a standard driving model, the driving operation of the driver when the standard driving model has been selected in the past is stored as a sample. That is, this sample shows the relationship between information about a moving body when a moving body such as a pedestrian has been detected in the past and the driving operation of the driver after the moving body is detected, Moreover, the correspondence relationship between the model determined to be used and the driving operation of the driver after the moving body is detected is shown. When the specified number of samples (for example, four times) can be acquired, the short-term fitness is calculated based on the stored number of samples. The goodness of fit is calculated by the following equation (2).
Goodness of fit = (N1 / Nt) × 100 (2)
Here, N1 is the number of samples outside the high risk area, and Nt is the number of all samples.

  FIG. 12 is a diagram illustrating an example of the fitness calculation. In FIG. 12, the fitness is calculated from a total of four samples, one sample in the high risk region R2 and three samples outside the high risk region R2. In this case, the fitness is calculated as 75% from the above equation (2). When the fitness is calculated, the process proceeds to step S202.

  In step S202, the model update determination unit 14 determines whether or not the fitness is greater than a certain value. The determination threshold value in step S202 is a reference value for determining whether or not the model conforms to the driving characteristics of the driver, and is set to 80%, for example. As a result of the determination in step S202, if it is determined that the fitness level is equal to or greater than the threshold (step S202-Y), the process proceeds to step S203, and if not (step S202-N), the process proceeds to step S204.

  In step S203, the model update determination unit 14 shifts from model update to driving behavior prediction processing. When step S203 is executed, the control flow ends.

  In step S204, the model update determination unit 14 shifts the model in which the risk area is small to the extent possible. FIG. 13 is a diagram illustrating an example of model shift by the model update determination unit 14. As shown in FIG. 13, the high-risk areas R11, R21, R31 after the shift are smaller than the high-risk areas R1, R2, R3 before the shift, respectively. In one shift, for example, the reference areas S1, S2, and S3 are shifted to the origin side so that the high risk areas R1, R2, and R3 are reduced by a certain amount or a certain ratio. As an example, a shift is made to decrease the maximum value of the deceleration rate of the high risk regions R1, R2, R3 at a constant rate with respect to each predicted side passage distance.

  For example, a driver with high skill may be too wide in the default high risk areas R1, R2, and R3, and the selected model may not match the driving characteristics of the driver. A driver with high skill can observe the behavior of the pedestrian 42 and appropriately perform avoidance behavior without slowing down so much. That is, even if the deceleration rate is the high-risk regions R1, R2, and R3 in the default model, it may be a deceleration rate that should be classified into the reference regions S1, S2, and S3 for the driver. If driving assistance based on the default model is given to such a driver, there is a risk that the assistance may be troublesome for the driver. On the other hand, the high risk areas R11, R21, R31 can be updated to more appropriate ones by shifting the model based on the degree of fitness calculated from the driving operation of the driver. Thereby, driving assistance according to the driver's needs becomes possible.

  It is preferable that the short-term update is repeatedly executed until the fitness level becomes equal to or higher than a threshold value. When the fitness becomes equal to or greater than the threshold value due to the short-term update, the short-term update for the model ends. Here, in the long term, the driving characteristics of the driver may change. For example, driving characteristics may change due to improvement in skill or familiarity with the vehicle, and there is a risk that the degree of fitness of the model will decrease. On the other hand, in this embodiment, long-term update for the model is executed. In the long-term update, the long-term fitness is calculated based on a sample of a specified period. The sample for which the long-term fitness is calculated may be all samples in a specified period, a sample in the latest fixed period, or a sample in the latest specified number of times. If the long-term fitness is less than the threshold, the model is shifted as in the case of short-term updates. By the long-term update, the degree of support is updated according to changes in the driving situation of the driver. Therefore, the driver can use the driving support technology for a long time.

  It should be noted that when the model is shifted, it is desirable that a certain restriction be imposed on the shift. For example, when driving assistance is performed by voice or video, it is necessary to ensure that the driver has enough time to move to avoiding behavior after providing assistance. Therefore, it is desirable to determine a minimum area to be secured in the high risk areas R11, R21, and R31 after the shift. When the model is shifted in step S204, the process proceeds to step S201.

  The model may be updated when the deceleration due to the driving operation of the driver deviates from the low risk areas T1, T2, and T3. In this case, in the above formula (2) for calculating the fitness, N1 may be the number of samples outside the low risk area. If the fitness is not greater than or equal to the threshold value, the reference areas S1, S2, and S3 are shifted to the opposite side to the origin side so as to reduce the low-risk areas T1, T2, and T3. By updating the model in this manner, it is possible to appropriately provide driving assistance to a driver who tends to decelerate significantly when there is a pedestrian ahead when he deviates from the normal deceleration operation. That is, the model can be updated to reduce the risk according to the driving characteristics of the driver.

  The model determination unit 15 determines a model used for control. The model determination unit 15 determines a driving model based on the update result by the model update determination unit 14 and the information collected by the own vehicle information collection unit 12. For example, when the model is updated by the model update determination unit 14, the updated model is selected as the model for determining support in place of the model before the update.

  The driving behavior prediction unit 16 includes a side passage distance prediction unit 16a and a deceleration rate calculation unit 16b. The side passage distance prediction unit 16a calculates the side passage prediction distance at the time of the measurement trigger (second point P1). The predicted side passage distance can be calculated based on the calculation result of the object information calculation unit 10 and the information collected by the own vehicle information collection unit 12. The deceleration rate calculation unit 16b calculates the reference vehicle speed V0 and the minimum vehicle speed V1 from the speeds detected by the vehicle information collection unit 12, and calculates the deceleration rate by the above equation (1).

  The driving behavior prediction determination unit 17 calculates the degree of deviation from the driving operation reference. FIG. 14 is an explanatory diagram of the degree of departure and the degree of departure recognition. In FIG. 14, the downward vertical axis indicates the departure degree, and the left horizontal axis indicates the driver departure recognition degree. The deviation degree indicates the degree to which the actual deceleration rate due to the driving operation of the driver deviates from the reference region S2. If the deceleration rate due to the driving operation is a value within the reference region S2 with respect to the calculated predicted side passage distance, the degree of deviation is zero. On the other hand, if the deceleration rate due to the driving operation is a value outside the reference region S2, the deviation degree is calculated as a value other than 0, and the deviation degree increases as the deceleration rate due to the driving operation deviates from the value in the reference region S2. Will increase.

  The magnitude of the departure degree is calculated in units of the width of the reference area S2. As shown in FIG. 14, one unit of the deviation degree is a difference between the maximum value and the minimum value of the reference area S2 in the calculated side passage predicted distance, that is, the width of the reference area S2 in the vertical axis direction. When the deceleration rate due to the driving operation is a value within the high-risk region R2, the difference between the deceleration rate value on the high-risk side boundary line H2 and the deceleration rate value due to the driving operation is divided by one unit of deviation. The thing is the deviation.

  The departure degree may be calculated when the deceleration rate due to the driving operation is a value within the low risk region T2. In this case, the deviation is obtained by dividing the difference between the value of the deceleration rate by the driving operation and the value of the deceleration rate on the low risk side boundary line L2 by one unit of the deviation degree. The deviation degree when the deceleration rate due to the driving operation is a value within the low risk region T2 may be a negative value.

  The support determination unit 18 determines whether or not to perform driving support based on the degree of deviation and a support level when driving support is performed. The driving support includes awareness support that transmits information to the driver by voice, light, video, vibration, and the like, and vehicle control support that supports avoidance behavior by controlling the host vehicle 100. In the awareness support and the vehicle control support, a plurality of support levels having different degrees of stimulation and the degree of intervention by control can be set.

  The correspondence relationship between the degree of departure and the support level can be determined in advance by, for example, the method described below. In FIG. 14, a broken line 300 indicates a distribution function (probability density function) obtained as a result of sensory evaluation, and a solid line 301 indicates a probability distribution function. The distribution function 300 is created based on the result of the psychological questionnaire. In the psychological questionnaire, the degree of deviation for each of the plurality of drivers that begins to realize that each driver has deviated from the driving operation in the reference region S2 is investigated. With the deviation of the median value of the distribution function 300, half of the drivers are aware that they have deviated from the reference area S2.

  The probability distribution function 301 is a curve obtained by integrating the distribution function 300. The probability distribution function 301 is a psychological deviation curve representing the driver's sensory deviation recognition degree. The support level is determined according to the probability distribution function 301, for example. As the probability distribution function 301 increases, there is a high possibility that the driver is aware of a deviation from the reference region S2. In other words, when the calculated deviation degree is a deviation degree corresponding to a large value of the probability distribution function 301, the driver is driving without noticing the presence of the pedestrian 42, or the pedestrian 42 is visually recognized. However, there is a high possibility of driving without paying sufficient attention. That is, if the probability distribution function 301 is a large value, there is a high possibility that the driving assistance is accepted by the driver. Moreover, it can be said that the higher the probability distribution function 301 is, the more preferable the driving assistance at a higher assistance level is.

  Therefore, by determining whether or not to perform driving support based on the value of the probability distribution function 301, and determining the support level when performing driving support, the driver's notice delay is suppressed, and the driver feels uncomfortable. It is possible to provide appropriate driving assistance that is difficult to give. In addition, the support level increases according to the size of the probability distribution function 301, so that the driver can intuitively understand the amount of deviation of the driving operation from the reference driving operation. As a result, the driver can drive You can feel the effectiveness of the support.

  The support determination unit 18 determines to perform vehicle control support in a situation where it is predicted that appropriate avoidance action by the driver is difficult in the awareness support. When the deceleration rate is small, the time until the host vehicle 100 approaches the pedestrian 42 is shortened. For this reason, if the driver starts the avoidance operation after recognizing the pedestrian 42 by the awareness support, the avoidance timing is delayed, and the risk may not be sufficiently reduced. The assistance determination unit 18 determines whether or not to perform vehicle assistance control based on, for example, the collision time TTC and the predicted lateral passage distance.

  The support determination unit 18 executes the determined driving support. The awareness support unit 19 controls the awareness device 30 based on the execution command of the awareness support by the support determination unit 18. The awareness device 30 is an information transmission device that transmits information to the driver by voice, light, video, vibration, and other stimuli. The awareness device 30 can transmit information at a plurality of support levels with different levels of stimulation. For example, when information is transmitted to the driver by a buzzer sound, a louder sound may be emitted as the support level is higher, or the interval between intermittent sounds may be shortened.

  The vehicle control support unit 20 executes vehicle control support based on a vehicle control support execution command from the support determination unit 18. The vehicle control support unit 20 can control the prime mover, the brake device, the steering device, and the like, and can assist the driver's driving operation, for example, the operation of avoiding the approach to the pedestrian 42 by controlling them. it can.

  Here, with reference to FIG. 1, the flow of the driving assistance of this embodiment is demonstrated. The control flow shown in FIG. 1 is repeatedly executed during traveling, for example.

  First, in step S101, the model selection unit 13 selects a default model. The model selection unit 13 reads a default model stored in the model database 11. When step S101 is executed, the process proceeds to step S102.

  In step S102, environmental information and own vehicle information are measured. The object information calculation unit 10 acquires environment information including information about the pedestrian 42 and information about the own lane 40 based on the detection result of the outside environment sensor. The host vehicle information collection unit 12 acquires host vehicle information such as the position, speed, steering angle, and pedal operation amount of the host vehicle 100.

  Next, in step S103, it is determined whether or not the relative distance and relative speed between the pedestrian 42 and the host vehicle 100 are within the measurement range. This determination is made by the model selection unit 13, for example. The model selection unit 13 determines whether or not the host vehicle 100 is in an area between the first point P0 and the second point P1 based on the relative distance between the host vehicle 100 and the pedestrian 42. If it is not determined that the host vehicle 100 is in the region between the first point P0 and the second point P1, a negative determination is made in step S103. Further, the model selection unit 13 determines whether or not the relative speed between the host vehicle 100 and the pedestrian 42 at the first point P0 is not less than the minimum vehicle speed Vmin and not more than the maximum vehicle speed Vmax. If it is not determined that the relative speed is not less than the minimum vehicle speed Vmin and not more than the maximum vehicle speed Vmax, a negative determination is made in step S103.

  If the result of determination in step S103 is affirmative (step S103-Y), the process proceeds to step S104. If not (step S103-N), the process proceeds to step S102.

  In step S104, the host vehicle information collection unit 12 observes a deceleration rate, a pedal operation amount, and the like. The own vehicle information collection unit 12 calculates a deceleration rate based on the speed of the own vehicle 100. When step S104 is executed, the process proceeds to step S105.

  In step S105, the short-term update determination unit 14a determines whether data necessary for model update has been acquired. Whether the short-term update determination unit 14a has acquired a necessary number of samples for a model selection parameter, for example, a combination of the lateral distance between the pedestrian 42 and the own lane 40 and the orientation of the pedestrian 42 Determine whether or not. FIG. 15 is a diagram illustrating an example of the number of data necessary for model update.

  For the combination of the direction of the pedestrian 42 and the lateral distance from the pedestrian 42, the number of samples (numerator) that can be acquired and the number of samples of the measurement level (denominator) necessary for model update are stored. In FIG. 15, data of the necessary number of samples can be acquired for a scene in which the direction of the pedestrian 42 is the own lane 40 side and the distance from the own lane 40 to the pedestrian 42 is within a certain distance. For other scenes, the number of samples is insufficient and the model cannot be updated yet. In this case, if the currently selected model is an updatable model, an update process is performed, otherwise the default model is used as it is.

  If data of the necessary number of samples has been acquired for the scene corresponding to the environmental information acquired in step S102, an affirmative determination is made in step S105. As a result of the determination in step S105, if it is determined that the data necessary for updating the model has been acquired (step S105-Y), the process proceeds to step S106. If not (step S105-N), the process proceeds to step S109. .

  In step S106, model update is determined by the short-term update determination unit 14a, and model update is executed. The short-term update determination unit 14a updates the model so that the fitness of the model satisfies a predetermined standard. When step S106 is executed, the process proceeds to step S107.

  In step S107, the long-term update determination unit 16b determines whether or not there is a need for long-term update for a model (update model) that has been short-term updated. The long-term update determination unit 16b calculates the long-term suitability of the current model (update model) based on observation results for a certain period such as monthly or yearly, and determines whether to update the model. As a result of the determination in step S107, if it is determined that the model needs to be updated (step S107-Y), the process proceeds to step S108, and if not (step S107-N), the process proceeds to step S110.

  In step S108, the degree of deviation from the re-update model is calculated. The long-term update determination unit 16b performs long-term update (re-update) on the update model corresponding to the current scene. The driving behavior prediction determination unit 17 calculates the deviation degree based on the re-update model that has been updated for a long time and the predicted side-passing distance and the deceleration rate calculated by the driving behavior prediction unit 16. When step S108 is executed, the process proceeds to step S111.

  In step S110, the degree of deviation from the update model is calculated. The driving behavior prediction determination unit 17 calculates the departure degree based on the updated model that has been updated in a short-term and the predicted side passage distance and the deceleration rate calculated by the driving behavior prediction unit 16. When step S110 is executed, the process proceeds to step S111.

  When a negative determination is made in step S105 and the process proceeds to step S109, the deviation from the default model is calculated in step S109. The driving behavior prediction determination unit 17 calculates the departure degree based on the default model and the predicted side passage distance and the deceleration rate calculated by the driving behavior prediction unit 16. When step S109 is executed, the process proceeds to step S111.

  In step S111, the support determination unit 18 determines whether or not the departure degree is large. The support determination unit 18 determines whether or not the deviation degree calculated in steps S108, S109, and S110 is large. For example, the support determination unit 18 performs the determination in step S111 based on the comparison result between the determination value determined based on the probability distribution function 301 and the calculated degree of deviation. As a result of the determination in step S111, if it is determined that the degree of departure is large (step S111-Y), the process proceeds to step S113. If not (step S111-N), the process proceeds to step S112.

  In step S112, the support determination unit 18 determines not to perform notification support. The support determination unit 18 outputs a command to turn off information provision by the notice device 30. The vehicle control support is also turned off because the degree of departure does not require awareness support. When step S112 is executed, this control flow ends.

  In step S113, the support determination unit 18 determines the execution of notification support. The support determination unit 18 outputs a command to turn on information provision by the notice device 30. The awareness support unit 19 controls the awareness device 30 in response to an information provision ON command, and performs driving support by notification. When step S113 is executed, the control flow ends.

  As described above, the driving support apparatus 1-1 according to the present embodiment corresponds to the correspondence between the information about the relative position between the moving object such as a pedestrian and the own vehicle detected around the own vehicle and the driving operation of the driver. A plurality of model candidates that define the relationship are provided, and a model to be used is determined from the plurality of model candidates based on information on the detected moving body, and the determined model and driving of the driver after the moving body is detected Driving assistance is executed based on the operation. Therefore, the necessity of driving assistance and the driving assistance level can be determined based on the driver's reaction to pedestrians and the like. Therefore, the driving assistance apparatus 1-1 has an advantage that driving assistance can be performed while suppressing the driver from feeling uncomfortable.

  In addition, the driving assistance apparatus 1-1 performs driving assistance when the deviation from the selected model is large, and changes the driving assistance level according to the degree of deviation. On the other hand, when the deviation from the model is small, driving assistance is not performed. That is, the driving assistance by the driving assistance apparatus 1-1 is based on the degree of deviation between the driving operation of the driver after the detection of a moving body such as a pedestrian and the driving operation of the selected model. Therefore, the driving assistance apparatus 1-1 can perform driving assistance that matches the driver's feeling.

  Each operation model such as a tension operation model, a standard operation model, and a margin operation model has different reference regions S1, S2, S3 and high risk regions R1, R2, R3. Therefore, the support level is determined within a range determined according to the selected model. In other words, the support level is determined based on the driving operation of the driver within a range determined according to the information about the moving body such as the pedestrian 42. Therefore, it is possible to determine the support level within an appropriate range according to the posture or movement of the pedestrian and the like, and it is possible to support in accordance with the driver's feeling.

  Further, in the driving support device 1-1 of the present embodiment, the necessity and level of driving support are determined based on the correspondence between the deviation from the reference areas S1, S2, and S3 and the driver deviation recognition degree, respectively. The Therefore, driving assistance that matches the driver's sense and does not give a sense of incongruity is performed.

  In addition, when the pedestrian 42 is going to cross the front of the own vehicle 100 or to start the crossing, the front crossing driving model shown in FIG. 16 can be used instead of the driving model shown in FIGS. . As shown in FIG. 16, the high risk region R4 of the forward crossing driving model extends to a region with a higher deceleration rate than the high risk regions R1, R2, and R3 of the other driving models. In other words, the reference region S4 when the predicted lateral passage distance is small is narrower than the reference regions S1, S2, S3 of the other driving models. In other words, when the pedestrian 42 is near the own lane 100 and is going to cross the lane, it is high if the vehicle is not decelerating close to 100% (so that it stops) until the second point P1. It will be judged as a risk and driving assistance will be started.

  In the present embodiment, the support level is determined based on the driving operation of the driver after the pedestrian is detected, but the timing for determining the support level is not limited to this. For example, a support level may be determined based on information about a pedestrian detected ahead, and the support level may be updated based on the driving operation of the driver. For example, when the tension driving model is selected, the relatively highest support level is set, the medium support level is set for the standard driving model, and the lowest support level is included for the margin driving model (including no support). Can be set. When the deviation degree of the driving operation of the driver is high, the driving support level is updated so that the risk is lower, and when the deviation degree is not high, the assistance level may not be updated. The support start may be executed after determining whether or not the support level needs to be updated based on the driving operation of the driver, for example.

  In this way, when the risk at the driving support level determined based on the posture and movement of the pedestrian is high, the driving support level is updated to lower the risk, while the risk is high. If not, the driving support level is not updated. Therefore, driving assistance considering the driver's reaction to the posture and movement of the pedestrian is possible. As a result, the driver is prevented from feeling uncomfortable with the support content.

[Modification of Embodiment]
A modification of the embodiment will be described. In the above embodiment, the deceleration rate is used as a driving operation for determining the degree of risk, but the driving operation is not limited to this. It is possible to calculate the degree of risk based on various detection results related to the driving operation such as the amount of driving operation of the driver, the operation timing, the operating force, the operation speed, or the vehicle behavior generated as a result of the driving operation.

  FIG. 17 is a diagram illustrating an example of an operation model in which the vertical axis indicates the operation timing. In the vertical axis, the proximity to the origin indicates that the operation timing is late, and the distance from the origin indicates that the operation timing is early. The side with the operation timing later than the operation timing of the reference region S5 is the high risk region R5, and the side with the operation timing earlier than the operation timing of the reference region S5 is the low risk region T5.

  The operation timing can be, for example, an accelerator-off timing or a brake-on timing. Not only this but the steering operation timing of the direction which avoids the pedestrian 42 may be made into the operation timing of FIG. The operation timing can be detected earlier than the vehicle behavior. Therefore, it is possible to determine the necessity of driving support and the support level at an early stage by evaluating the risk according to the operation timing. In addition, when detecting the timing of driving operation or the like instead of the vehicle behavior, the influence of the disturbance is difficult to enter, and the driver's reaction can be directly detected.

  The contents disclosed in the above embodiments and modifications can be implemented in appropriate combination.

1-1 Driving support device 40 Own lane 41 White line 42 Pedestrian 100 Own vehicle V0 Reference own vehicle speed V1 Minimum own vehicle speed P0 First point P1 Second point H1, H2, H3 High risk side boundary line L1, L2, L3 Low risk side boundary line R1, R2, R3 High risk area S1, S2, S3 Reference area T1, T2, T3 Low risk area

Claims (5)

A plurality of model candidates that define the correspondence relationship between the information on the relative position between the moving body detected in the vicinity of the host vehicle and the host vehicle, and the driving operation of the driver,
Determining a model to be used from the plurality of model candidates based on the information on the detected moving object;
Driving assistance is performed based on the determined model and the driving operation of the driver after the moving body is detected. The driving support device according to claim 1, wherein the determined model can be updated based on the determined model and the driving operation of the driver after the moving body is detected. The degree of fit between the determined model and the driving operation of the driver after the moving body is detected is calculated based on the correspondence between the model and the driving operation for a predetermined number of times, and the degree of fit is The driving support device according to claim 2, wherein the update is executed when the value is less than a set reference value. The driving support device according to claim 2, wherein the determined model is updated according to a short-term fitness and a long-term fitness with the driving operation of the driver after the moving body is detected. The driving assistance apparatus according to claim 1 or 2, wherein the driving assistance is based on a degree of deviation between the determined model and the driving operation of the driver after the moving body is detected.

Images